Research map / Learning theory and algorithms / Kernels and regression theory
On the Optimal Weighted $\ell_2$ Regularization in Overparameterized Linear Regression
Denny Wu, Ji Xu
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Use a negative ridge parameter when feature variance and signal strength are positively aligned in the noiseless overparameterized regime | Standard ridge regression · Overparameterized linear regression, noiseless | prediction risk | Improved |
| Use a positive ridge parameter when feature variance and signal strength are negatively aligned | Standard ridge regression · Overparameterized linear regression, noiseless | prediction risk | Improved |
| Increase the noise level in a setting where the optimal ridge parameter is negative | Overparameterized linear regression · Linear regression with anisotropic features and signal | optimal ridge parameter | Mixed |
| Use positive rather than negative ridge regularization | Ridge regression · Underparameterized regime, γ<1 | prediction risk | Improved |
| Increase the retained principal-component fraction | Principal component regression · Overparameterized regime, θγ>1, with decreasing E[g|h] | prediction risk | Improved |
| Use weighting matrix Σ_w=Σ_x rather than another admissible weighting | Ridgeless generalized ridge regression · Overparameterized linear regression | variance | Improved |
| Use weighting matrix Σ_w=Σ̄_β^{-1} rather than another admissible weighting | Ridgeless generalized ridge regression · Overparameterized linear regression | bias | Improved |
| Use weighting matrix Σ_w=Σ̄_β^{-1} with optimally tuned λ | Optimally tuned generalized ridge regression · Overparameterized linear regression | prediction risk | Improved |
| Use a weighting matrix based only on Σ_x, proportional to (f_v(Σ_x))^{-1} | Optimally tuned generalized ridge regression · Overparameterized linear regression when signal covariance is unknown | prediction risk | Improved |
| Use the minimum weighted-norm solution instead of PCR | Ridgeless generalized ridge regression with Σ_w=Σ̄_β^{-1} or (f_v(Σ_x))^{-1} · Noiseless overparameterized linear regression | prediction risk | Improved |
| Increase the overparameterization ratio γ | Optimally tuned standard ridge regression with isotropic signal prior · Linear regression with general data covariance | prediction risk | Hurt |
Closest papers by meaning
- Benign overfitting in ridge regression (JMLR 2023 · similarity 0.89)
- The Distribution of Ridgeless Least Squares Interpolators (JMLR 2026 · similarity 0.88)
- Bayes beats Cross Validation: Efficient and Accurate Ridge Regression via Expectation Maximization (NeurIPS 2023 · similarity 0.87)
- Risk Bounds of Accelerated SGD for Overparameterized Linear Regression (ICLR 2024 · similarity 0.86)
- Regularized least squares learning with heavy-tailed noise is minimax optimal (NeurIPS 2025 · similarity 0.85)
- Characterizing Overfitting in Kernel Ridgeless Regression Through the Eigenspectrum (ICML 2024 · similarity 0.85)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
